The Art of Research with NotebookLM and Readwise

I I recently conducted a research on the state of AI agent solutions on the market, which is growing every day and has become the de facto trend in the AI space at the moment. For this purpose, there are so many ways to do an AI-powered research like this:

• Using ChatGPT web search
• Using Perplexity
• Using Gpt-Researcher with Tavily
• And a few more.

They are great for getting a good overview of research material relevant to your topic. But the reality is sometimes a little different: Depending on the context, you may only want to consider curatedreliable sources for your research, depending on factors like trust, compliance, etc.

For example, if I were comparing two hosting providers such as Vercel and Railway, I would not necessarily rely on technical details provided by a third party, I would prefer sources directly from the provider, or at least consider them more relevant.

Approach to Curated Research

For a more fine-tuned approach (taking AI agents as an example), I’d usually do something like this:

1. Do a general overview search to find suitable sources (e.g., Wikipedia, Hugging Face, LinkedIn posts).
2. Check these sources and pick out the relevant parts by reading through them and selecting where I want to dig deeper—like choosing a list of open-source agent solutions I might explore.
3. Check the vendor or GitHub repositories of those solutions directly.

Sometimes the sources aren’t just websites. They could be YouTube videos, PDF files, and so on. We have to consider different formats accordingly.

In the past, I often didn’t go beyond the initial “overview” stage of research. Most of the deeper exploration was done manually.

Introducing Readwise and NotebookLM

I’m a big fan of Readwise, so I decided to use it to do more fine-grained research, especially now that it supports NotebookLM.

What is NotebookLM?
NotebookLM is a conversational AI tool developed by Google that helps users organize, explore, and interact with their notes. It leverages large language models to summarize, search, and respond to questions about the content within notebooks. Designed for knowledge management, it allows users to manage information more efficiently. The tool aims to improve productivity by integrating AI with personal or collaborative note-taking.

What is Readwise?
Readwise is a tool designed to help users save, organize, and revisit highlights from books, articles, and other reading materials. It syncs highlights from platforms like Kindle, Pocket, and Instapaper, and sends daily emails to remind users of their saved highlights. The service also offers features to organize and categorize notes for easier access. Readwise aims to improve retention and learning by helping users revisit important information regularly.

How It Works

1. Do your usual research and choose the right sources using GPT, Perplexity, etc.
2. Check the articles (PDF, text, video, whatever). If they have something useful, highlight those relevant parts. Be sure to tag your highlights so you can focus on them later. For AI agents, I used the tag “ai-agents.”
3. Sometimes, dive deeper into more sources as described above. Highlight the interesting parts, like product features.

After saving everything in Readwise, you can do all sorts of stuff—make a brief, test your knowledge (quiz), write documentation, or even create a podcast using NotebookLM.

Integrating NotebookLM with Readwise

1. Open the export dashboard and select NotebookLM. Connect it:
   https://readwise.io/export

1. Deselect all articles and enter the tag you chose above (e.g., ai-agent, without the #). Then select all from the search results.

After the export finishes, you should see the exported articles in your Google Drive.

1. Once exported to Google Drive, head to NotebookLM, select Google Drive as the source, and import the files you just exported.

Now you can enjoy the combined power of Readwise and NotebookLM.

You’ll see an automatically generated summary of the content from NotebookLM:

You can now generate FAQs, study guides, or even produce a podcast from your curated materials. This approach makes absorbing the material much easier:

You can also use custom prompts and the built-in chat function for quizzes:

Make a quiz with one question at a time and wait for my answer before showing the answer.

This approach has really helped me conduct more reliable research based on curated information. I believe it can help you too, whether you’re studying, doing research, or engaging in similar activities.